Project Summary Multiple Myeloma (MM) is an incurable malignancy of plasma cells affecting more than 20,000 patients each year, with a median survival of approximately 6 years. Despite recent advancements in therapy, the disease remains fatal in the majority of patients. Genetic aberrations including translocations, mutations in crucial cellular pathways, and copy number alterations, have been identified as molecular hallmarks of MM. However, the causal drivers of MM pathogenesis are still unclear, and treatment is given empirically based on recurrence risk rather than genetic events. High-throughput DNA sequencing of MM patients has revealed wide and remarkable heterogeneity of the mutational spectrum across the cohort of patients and complex sub-clonal structure. Clonal heterogeneity, evolution and competition suggest that the employment of a personalized therapeutic approach is likely to improve the outcomes for myeloma. In the past recent years, integrative network-based methods have proven to be extremely effective in modeling complex biological systems and uncovering novel patterns of genomic perturbation associated to disease initiation and progression. However, to date, there have been no network-based models in hematological malignancies developed using large-scale next generation sequencing data. In on going research, we have developed an integrative network biology approach to infer an improved molecular model of newly diagnosed MM. Our preliminary results show that our methodology is able to successfully recover previous knowledge and identify novel biological features of MM at both general and patient-specific level. This makes our approach naturally suitable for personalized medicine applications. Moreover, case studies of patients treated at Mount Sinai based on deep sequencing analysis show that a data-driven personalized therapy approach could be an effective way of treating patients resistant to the standard of care. To take our methodology to the next level, we propose to develop a complete ?bench- to-bedside? intelligent learning precision medicine platform for personalized cancer therapy, specifically designed for MM based on large datasets and a novel comprehensive integrative network approach. We will generate network models of myeloma patients in order to infer key driver genetic alterations, derive prognostic signatures and stage-specific features, perform disease subtype classification and inform therapy prioritization. The tool will match individual patients profiles with network-inferred disease classes, in order to select the most appropriate treatments, and will perform combination therapy prediction using tumor clonality assessment. Physician's opinion and therapy outcome will be used as feedback to refine the predictions through reinforcement learning techniques. It is important to note that, although the system will be built based on MM models, it will be practically applicable to other types of cancer, providing a general framework for personalized network-based therapy in oncology.